Elamur Research⌕
☰
home / papers / 2025 / ChemAgent: Self-updating Library in Large Language Models Improves Chemical Reasoning

ChemAgent: Self-updating Library in Large Language Models Improves Chemical Reasoning

Paper · 2025 · Xiangru Tang, Tianyu Hu, Muyang Ye et al.

Summary

Large language models often fail at multi-step chemistry problems. They misuse domain formulas, choose wrong constants or units, and make arithmetic slips that carry through later steps. They also write faulty Python for calculations. Earlier fixes such as StructChem prompting or self-reflection rely on fixed workflows or hand-curated knowledge, and they do not learn from previously solved problems. ChemAgent addresses this with an external library that updates itself. Problems from a development set are split into small sub-tasks, each with its conditions and a solution. These are verified and stored, and incorrect entries are dropped. At test time, a new problem is decomposed in the same way. Related stored strategies and worked sub-tasks are retrieved to guide each step, sub-solutions are checked and corrected, and newly validated material is added back to the library. Tested on four chemistry datasets from SciBench with GPT-3.5, GPT-4 and Llama3, the framework substantially raises accuracy and outperforms StructChem. The results suggest that accumulated, structured experience makes LLM scientific reasoning more reliable, with possible uses in drug discovery and materials research.

Key points

  • The library holds three memory types:
    • Planning Memory (M_p) stores high-level strategies.
    • Execution Memory (M_e) stores structured condition, sub-task and solution records.
    • Knowledge Memory (M_k) stores formulas and principles. It is generated on the fly for each task and not kept permanently.
  • The pipeline has two stages:
    • Initial library construction on the dev set, which builds long-term M_p and M_e.
    • Inference with memory-enhanced reasoning, evaluation and refinement, and dynamic library updates.
  • With GPT-4, ChemAgent improves accuracy over the base model by 37% on average and by up to 46%.
  • Against the previous state of the art, StructChem, it gains about 10% on average and up to 15% across datasets.
  • Stronger base models benefit more: GPT-4 improves more than GPT-3.5. Code is available at github.com/gersteinlab/chemagent.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Stanford UniversityStanford · United States434 works · 314,419 citations🎓University of Illinois Urbana-ChampaignUrbana · United States184 works · 69,372 citations🎓Shanghai Jiao Tong UniversityShanghai · China149 works · 16,161 citations🎓Yale UniversityNew Haven · United States95 works · 12,765 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksComputer Vision and Pattern Recognition · 2,869 worksComputational Theory and Mathematics · 252 worksMaterials Science · 26 worksMaterials Chemistry · 26 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 works

Cite

Cite as Xiangru Tang, Tianyu Hu, Muyang Ye, Yanjun Shao, Xunjian Yin, Siru Ouyang, Wangchunshu Zhou, Pan Lu, Zhuosheng Zhang, Yilun Zhao, Arman Cohan and Mark Gerstein. 2025. ChemAgent: Self-updating Library in Large Language Models Improves Chemical Reasoning. arXiv:2501.06590.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2501.06590 · arXiv 2501.06590

@misc{tang2025chemagent,
  title = {{ChemAgent: Self-updating Library in Large Language Models Improves Chemical Reasoning}},
  author = {Xiangru Tang, Tianyu Hu, Muyang Ye, Yanjun Shao, Xunjian Yin, Siru Ouyang, Wangchunshu Zhou, Pan Lu, Zhuosheng Zhang, Yilun Zhao, Arman Cohan and Mark Gerstein},
  year = {2025},
  doi = {10.48550/arxiv.2501.06590},
  eprint = {2501.06590},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2501.06590}
}

References

67 references, 23 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Peter Atkins, Julio De Paula, and Ronald Friedman. Physical chemistry: quanta, matter, and change. Oxford University Press, USA, 2014. find
  2. Peter Atkins, Julio De Paula, and James Keeler. Atkins’ physical chemistry. Oxford university press, 2023. find
  3. Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th annual international conference on machine learning, pp. 41–48, 2009.
  4. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, et al. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on Artificial Int… find
  5. Daniil A. Boiko, Robert MacKnight, Ben Kline, and Gabe Gomes. Autonomous chemical research with large language models. Nature, 624:570 – 578, 2023.
  6. Andres M Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D White, and Philippe Schwaller. ChemCrow: Augmenting large-language models with chemistry tools. arXiv preprint arXiv:2304.05376, 2023. find
  7. He Cao, Yanjun Shao, Zhiyuan Liu, Zijing Liu, Xiangru Tang, Yuan Yao, and Yu Li. PRESTO: Progressive pretraining enhances synthetic chemistry outcomes. arXiv preprint arXiv:2406.13193, 2024. find
  8. Aaron Chan, Zhiyuan Zeng, Wyatt Lake, Brihi Joshi, Hanjie Chen, and Xiang Ren. KNIFE: Distilling meta-reasoning knowledge with free-text rationales. In ICLR 2023 Workshop on Pitfalls of limited data and computation for Trustworthy ML, 2023. find
  9. Kourosh Darvish, Marta Skreta, Yuchi Zhao, Naruki Yoshikawa, Sagnik Som, Miroslav Bogdanovic, Yang Cao, Han Hao, Haoping Xu, Alán Aspuru-Guzik, Animesh Garg, and Florian Shkurti. Organa: A robotic assistant for automated chemistry experimentation and characterization. ArXiv, abs/2401.06949, 2024. — [arXiv:2401.06949](https://arxiv.org/abs/2401.06949) find
  10. Ernest Davis and Scott Aaronson. Testing GPT-4 with wolfram alpha and code interpreter plug-ins on math and science problems. arXiv preprint arXiv:2308.05713, 2023. find
  11. Kehua Feng, Keyan Ding, Weijie Wang, Xiang Zhuang, Zeyuan Wang, Ming Qin, Yu Zhao, Jianhua Yao, Qiang Zhang, and Huajun Chen. SciKnowEval: Evaluating multi-level scientific knowledge of large language models. arXiv preprint arXiv:2406.09098, 2024.
  12. Chrisantha Fernando, Dylan Sunil Banarse, Henryk Michalewski, Simon Osindero, and Tim Rocktäschel. Promptbreeder: Self-referential self-improvement via prompt evolution. In Forty-first International Conference on Machine Learning, 2023.
  13. Yao Fu, Hao Peng, Ashish Sabharwal, Peter Clark, and Tushar Khot. Complexity-based prompting for multi-step reasoning. In The Eleventh International Conference on Learning Representations, 2022.
  14. Luyu Gao, Zhuyun Dai, Panupong Pasupat, Anthony Chen, Arun Tejasvi Chaganty, Yicheng Fan, Vincent Zhao, Ni Lao, Hongrae Lee, Da-Cheng Juan, et al. Rarr: Researching and revising what language models say, using language models. In Proceedings of the 61st Annual Meeting of the Association for Computat…
  15. Zhibin Gou, Zhihong Shao, Yeyun Gong, Yujiu Yang, Nan Duan, Weizhu Chen, et al. Critic: Large language models can self-correct with tool-interactive critiquing. In The Twelfth International Conference on Learning Representations, 2023.
  16. Jing Guo, Nan Li, Jian Qiang Qi, Hang Yang, Ruiqiao Li, Yuzhen Feng, Si Zhang, and Ming Xu. Empowering working memory for large language model agents. ArXiv, abs/2312.17259, 2023a. — [arXiv:2312.17259](https://arxiv.org/abs/2312.17259) find
  17. Taicheng Guo, Bozhao Nan, Zhenwen Liang, Zhichun Guo, Nitesh Chawla, Olaf Wiest, Xiangliang Zhang, et al. What can large language models do in chemistry? a comprehensive benchmark on eight tasks. Advances in Neural Information Processing Systems, 36:59662–59688, 2023b. find
  18. Joseph F Hair Jr, Wiliam C Black, Barry J Babin, and Rolph E Anderson. Multivariate data analysis. In Multivariate data analysis, pp. 785–785. Prentice Hall, 2010. find
  19. Shengran Hu, Cong Lu, and Jeff Clune. Automated design of agentic systems. arXiv preprint arXiv:2408.08435, 2024. find
  20. Zhiting Hu and Tianmin Shu. Language models, agent models, and world models: The law for machine reasoning and planning. arXiv preprint arXiv:2312.05230, 2023. find
  21. Jie Huang, Xinyun Chen, Swaroop Mishra, Huaixiu Steven Zheng, Adams Wei Yu, Xinying Song, and Denny Zhou. Large language models cannot self-correct reasoning yet. In The Twelfth International Conference on Learning Representations, 2023.
  22. Xu Huang, Weiwen Liu, Xiaolong Chen, Xingmei Wang, Hao Wang, Defu Lian, Yasheng Wang, Ruiming Tang, and Enhong Chen. Understanding the planning of llm agents: A survey. ArXiv, abs/2402.02716, 2024a. find
  23. Yuqing Huang, Rongyang Zhang, Xuesong He, Xuyang Zhi, Hao Wang, Xin Li, Feiyang Xu, Deguang Liu, Huadong Liang, Yi Li, Jian Cui, Zimu Liu, Shijin Wang, Guoping Hu, Guiquan Liu, Qi Liu, Defu Lian, and Enhong Chen. Chemeval: A comprehensive multi-level chemical evaluation for large language models. ar… — [arXiv:2409.13989](https://arxiv.org/abs/2409.13989) find
  24. Shuyang Jiang, Yuhao Wang, and Yu Wang. Selfevolve: A code evolution framework via large language models. arXiv preprint arXiv:2306.02907, 2023. — [arXiv:2306.02907](https://arxiv.org/abs/2306.02907) find
  25. Zhanming Jie, Trung Quoc Luong, Xinbo Zhang, Xiaoran Jin, and Hang Li. Design of chain-of-thought in math problem solving. arXiv preprint arXiv:2309.11054, 2023. find
  26. Justin Johnson, Bharath Hariharan, Laurens Van Der Maaten, Li Fei-Fei, C Lawrence Zitnick, and Ross Girshick. Clevr: A diagnostic dataset for compositional language and elementary visual reasoning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2901–2910, 2017.
  27. Scott Barry Kaufman. Intelligence and the cognitive unconscious. The Cambridge handbook of intelligence, pp. 442–467, 2011. find
  28. Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. Decomposed prompting: A modular approach for solving complex tasks. In The Eleventh International Conference on Learning Representations, 2022. find
  29. Jiawei Li, Yizhe Yang, Yu Bai, Xiaofeng Zhou, Yinghao Li, Huashan Sun, Yuhang Liu, Xingpeng Si, Yuhao Ye, Yixiao Wu, et al. Fundamental capabilities of large language models and their applications in domain scenarios: A survey. In Proceedings of the 62nd Annual Meeting of the Association for Computa… find
  30. Kun Li, Xin Jing, and Chengang Jing. Vector storage based long-term memory research on llm. International Journal of Advanced Network, Monitoring and Controls, 2024b. find
  31. Zekun Li, Xianjun Yang, Kyuri Choi, Wanrong Zhu, Ryan Hsieh, HyeonJung Kim, Jin Hyuk Lim, Sungyoung Ji, Byungju Lee, Xifeng Yan, et al. MMSci: A multimodal multi-discipline dataset for PhD-level scientific comprehension. arXiv preprint arXiv:2407.04903, 2024c.
  32. Chang Liao, Yemin Yu, Yu Mei, and Ying Wei. From words to molecules: A survey of large language models in chemistry. arXiv preprint arXiv:2402.01439, 2024. — [arXiv:2402.01439](https://arxiv.org/abs/2402.01439) find
  33. AI@Meta Llama Team. The llama 3 herd of models, July 2024. URL https://llama.meta.com/. ↗
  34. Pan Lu, Swaroop Mishra, Tanglin Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. Learn to explain: Multimodal reasoning via thought chains for science question answering. Advances in Neural Information Processing Systems, 35:2507–2521, 2022. find
  35. Donald A McQuarrie. Quantum chemistry. University Science Books, 2008. find
  36. Adrian Mirza, Nawaf Alampara, Sreekanth Kunchapu, Benedict Emoekabu, Aswanth Krishnan, Mara Wilhelmi, Macjonathan Okereke, Juliane Eberhardt, Amir Mohammad Elahi, Maximilian Greiner, Caroline T. Holick, Tanya Gupta, Mehrdad Asgari, Christina Glaubitz, Lea C. Klepsch, Yannik Köster, Jakob Meyer, San… — [arXiv:2404.01475](https://arxiv.org/abs/2404.01475) find
  37. OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, et al. GPT-4 technical report, 2024.
  38. Magda Osman. An evaluation of dual-process theories of reasoning. Psychonomic bulletin & review, 11(6):988–1010, 2004. find
  39. Siru Ouyang, Zhuosheng Zhang, Bing Yan, Xuan Liu, Yejin Choi, Jiawei Han, and Lianhui Qin. Structured chemistry reasoning with large language models. In Forty-first International Conference on Machine Learning, 2024. find
  40. Pruthvi Patel, Swaroop Mishra, Mihir Parmar, and Chitta Baral. Is a question decomposition unit all we need? In 2022 Conference on Empirical Methods in Natural Language Processing, EMNLP 2022, 2022. find
  41. Qizhi Pei, Lijun Wu, Kaiyuan Gao, Jinhua Zhu, Yue Wang, Zun Wang, Tao Qin, and Rui Yan. Leveraging biomolecule and natural language through multi-modal learning: A survey. arXiv preprint arXiv:2403.01528, 2024. — [arXiv:2403.01528](https://arxiv.org/abs/2403.01528) find
  42. Cheng Qian, Shihao Liang, Yujia Qin, Yining Ye, Xin Cong, Yankai Lin, Yesai Wu, Zhiyuan Liu, and Maosong Sun. Investigate-consolidate-exploit: A general strategy for inter-task agent self-evolution. arXiv preprint arXiv:2401.13996, 2024. — [arXiv:2401.13996](https://arxiv.org/abs/2401.13996) find
  43. Yuxiao Qu, Tianjun Zhang, Naman Garg, and Aviral Kumar. Recursive introspection: Teaching language model agents how to self-improve. arXiv preprint arXiv:2407.18219, 2024. find
  44. Noah Shinn, Beck Labash, and Ashwin Gopinath. Reflexion: an autonomous agent with dynamic memory and self-reflection. arXiv preprint arXiv:2303.11366, 2023.
  45. Marta Skreta, Zihan Zhou, Jia Lin Yuan, Kourosh Darvish, Alán Aspuru-Guzik, and Animesh Garg. Replan: Robotic replanning with perception and language models. ArXiv, abs/2401.04157, 2024. — [arXiv:2401.04157](https://arxiv.org/abs/2401.04157) find
  46. Liangtai Sun, Yang Han, Zihan Zhao, Da Ma, Zhennan Shen, Baocai Chen, Lu Chen, and Kai Yu. SciEval: A multi-level large language model evaluation benchmark for scientific research. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38,17, pp. 19053–19061, 2024a. find
  47. Lichao Sun, Yue Huang, Haoran Wang, Siyuan Wu, Qihui Zhang, Chujie Gao, Yixin Huang, Wenhan Lyu, Yixuan Zhang, Xiner Li, et al. Trustllm: Trustworthiness in large language models. arXiv preprint arXiv:2401.05561, 2024b.
  48. Vicente Talanquer. The complexity of reasoning about and with chemical representations. JACS Au, 2(12):2658–2669, 2022. doi: 10.1021/jacsau.2c00498. — [doi:10.1021/jacsau.2c00498](https://doi.org/10.1021/jacsau.2c00498) find
  49. David Wadden, Kejian Shi, Jacob Morrison, Aakanksha Naik, Shruti Singh, Nitzan Barzilay, Kyle Lo, Tom Hope, Luca Soldaini, Shannon Zejiang Shen, et al. SciRIFF: A resource to enhance language model instruction-following over scientific literature. arXiv preprint arXiv:2406.07835, 2024.
  50. Xiaoxuan Wang, Ziniu Hu, Pan Lu, Yanqiao Zhu, Jieyu Zhang, Satyen Subramaniam, Arjun R Loomba, Shichang Zhang, Yizhou Sun, and Wei Wang. SciBench: Evaluating college-level scientific problem-solving abilities of large language models. In Forty-first International Conference on Machine Learning, 2024…
  51. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. In The Eleventh International Conference on Learning Representations, 2023.
  52. Zhiruo Wang, Jiayuan Mao, Daniel Fried, and Graham Neubig. Agent workflow memory. arXiv preprint arXiv:2409.07429, 2024b.
  53. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022.
  54. Kangda Wei, Dawn Lawrie, Benjamin Van Durme, Yunmo Chen, and Orion Weller. When do decompositions help for machine reading? In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pp. 3599–3606, 2023. find
  55. Yixuan Weng, Minjun Zhu, Shizhu He, Kang Liu, and Jun Zhao. Large language models are reasoners with self-verification. arXiv preprint arXiv:2212.09561, 2, 2022.
  56. Yi Xiao, Xiangxin Zhou, Qiang Liu, and Liang Wang. Bridging text and molecule: A survey on multimodal frameworks for molecule. arXiv preprint arXiv:2403.13830, 2024. — [arXiv:2403.13830](https://arxiv.org/abs/2403.13830) find
  57. Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V Le, Denny Zhou, and Xinyun Chen. Large language models as optimizers. In The Twelfth International Conference on Learning Representations, 2023.
  58. Ling Yang, Zhaochen Yu, Tianjun Zhang, Shiyi Cao, Minkai Xu, Wentao Zhang, Joseph E Gonzalez, and Bin Cui. Buffer of thoughts: Thought-augmented reasoning with large language models. arXiv preprint arXiv:2406.04271, 2024. find
  59. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36, 2024. find
  60. Yuan Zhang, Chao Wang, Juntong Qi, and Yan Peng. Leave it to large language models! correction and planning with memory integration. Cyborg and Bionic Systems, 5, 2023. find
  61. Yunxiang Zhang, Muhammad Khalifa, Lajanugen Logeswaran, Jaekyeom Kim, Moontae Lee, Honglak Lee, and Lu Wang. Small language models need strong verifiers to self-correct reasoning. arXiv preprint arXiv:2404.17140, 2024a. find
  62. Zeyu Zhang, Xiaohe Bo, Chen Ma, Rui Li, Xu Chen, Quanyu Dai, Jieming Zhu, Zhenhua Dong, and Ji-Rong Wen. A survey on the memory mechanism of large language model based agents. ArXiv, abs/2404.13501, 2024b. find
  63. Ming Zhong, Siru Ouyang, Minhao Jiang, Vivian Hu, Yizhu Jiao, Xuan Wang, and Jiawei Han. ReactIE: Enhancing chemical reaction extraction with weak supervision. In Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki (eds.), Findings of the Association for Computational Linguistics: ACL 2023, pp. 1212… find
  64. Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. Memorybank: Enhancing large language models with long-term memory. Proceedings of the AAAI Conference on Artificial Intelligence, 38(17):19724–19731, Mar. 2024a. doi: 10.1609/aaai.v38i17.29946. URL https://ojs.aaai.org/index.php/AAAI/ar… ↗
  65. Xianrui Zhong, Yufeng Du, Siru Ouyang, Ming Zhong, Tingfeng Luo, Qirong Ho, Hao Peng, Heng Ji, and Jiawei Han. Actionie: Action extraction from scientific literature with programming languages. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long… find
  66. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc V Le, and Ed H. Chi. Least-to-most prompting enables complex reasoning in large language models. In The Eleventh International Conference on Learning Representations, 20…
  67. Wangchunshu Zhou, Yixin Ou, Shengwei Ding, Long Li, Jialong Wu, Tiannan Wang, Jiamin Chen, Shuai Wang, Xiaohua Xu, Ningyu Zhang, et al. Symbolic learning enables self-evolving agents. arXiv preprint arXiv:2406.18532, 2024.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tang-2025-chemagent-self-updating-library-large/